Ultrasound Obstet Gynecol 2013; 41: 672–678
Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/uog.12423
Predicting successful vaginal birth after Cesarean section
using a model based on Cesarean scar features examined
by transvaginal sonography
O. NAJI*†, L. WYNANTS‡§, A. SMITH*, Y. ABDALLAH*†, C. STALDER*, A. SAYASNEH*†,
A. McINDOE*, S. GHAEM-MAGHAMI†, S. VAN HUFFEL‡§, B. VAN CALSTER¶,
D. TIMMERMAN¶ and T. BOURNE*†¶
*Obstetrics and Gynaecology Unit, Queen Charlottes and Chelsea Hospital, Imperial College, London, UK; †Institute of Reproductive and
Developmental Biology, Imperial College, London, UK; ‡Department of Electrical Engineering (ESAT-SCD), KU Leuven, Leuven, Belgium;
§iMinds Future Health Department, KU Leuven–University of Leuven, Leuven, Belgium; ¶Department of Development and Regeneration,
University Hospitals, KU Leuven, Belgium
K E Y W O R D S: Cesarean section scar; transvaginal sonography; vaginal birth after Cesarean section; vaginal delivery; VBAC
ABSTRACT
Objective To develop a model to predict the success of
a trial of vaginal birth after Cesarean section (VBAC)
based on sonographic measurements of Cesarean section
(CS) scar features, demographic variables and previous
obstetric history.
Methods We used transvaginal sonography (TVS) to
examine the CS scar of 320 consecutive pregnant women.
TVS was carried out at 11–13, 19–21 and 34–36 weeks’
gestation and prospective measurements of the scar were
recorded at each visit according to a defined protocol.
A logistic regression model to predict success of VBAC
was developed for those patients with a visible scar on
ultrasound and only one previous CS. The model was
evaluated using bootstrap validation.
Results There were 131 women with one previous CS and
a visible scar, of whom 10 underwent CS prior to labor
and were excluded from analysis. Successful VBAC was
achieved in 74/121 (61%) of the remaining cases. The
prediction model developed was based on patient age,
previous history of VBAC, residual myometrial thickness
(RMT) and the change in RMT from the first to the second
trimester (!RMT). The internally validated area under the
receiver–operating characteristics curve was 0.62 when
measurements of RMT and !RMT were excluded, but
0.94 when scar information was incorporated into the
model.
Conclusion Ultrasound measurements of CS scar, namely
RMT and the change in RMT from the first to the
second trimester of pregnancy, when incorporated into
a mathematical model, can predict accurately a successful
trial of labor in patients with one previous CS. Copyright
 2013 ISUOG. Published by John Wiley & Sons Ltd.
INTRODUCTION
The rate of vaginal birth after Cesarean section (VBAC)
is defined as the number of vaginal births to women
with one previous Cesarean section (CS) per 100 such
deliveries1 . Attempts have been made to predict which
patients are more likely to undergo successful VBAC using
various parameters, including clinical history and physical
examination at the time of admission for delivery2 . Flamm
and Geiger3 developed a scoring system to predict the
success of a trial of labor in these patients, in which
they allocated points for maternal age, a history of
previous vaginal delivery, the indication for the previous
CS delivery and cervical effacement and dilation. When
the system was tested, they found that 18% of women
scored ≤ 3, corresponding to a lower than 60% likelihood
of VBAC, whilst only 29% of women scored ≥ 6, leading
to an estimated 88% likelihood of VBAC. However, the
fact that higher scores in their model were applicable
to fewer than 30% of the study population limits its
utility. Dinsmoor and Brock4 tested the Flamm and Geiger
model retrospectively on 153 cases, finding that a score
of ≥ 7 gave a 100% likelihood of VBAC, whilst a score
of ≤ 4 reduced the chances to 53%. They concluded that
an unfavorable score might be helpful when counseling
patients considering a trial of labor. However, this system
does not give a calibrated spectrum of risk and leaves any
decision regarding VBAC until a woman is admitted to
Correspondence to: Dr S. Ghaem-Maghami, Institute of Development and Reproductive Biology IRDB, Imperial College London,
Hammersmith Campus, Du Cane Road, London W12 0HS, UK (e-mail: [email protected])
Accepted: 27 December 2012
Copyright  2013 ISUOG. Published by John Wiley & Sons Ltd.
ORIGINAL PAPER
673
TVS of Cesarean scar to predict successful vaginal birth
the delivery room in labor. A better system, able to predict
the success or failure of a trial of labor at an earlier stage
in pregnancy, is lacking.
Accurately predicting the outcome of a trial of VBAC
is clinically useful as failure is associated with increased
maternal and fetal morbidity5 . There is a growing body of
evidence to suggest that complete healing of the CS scar
and myometrial thickness of the lower uterine segment
are related to the chance of achieving a vaginal delivery
in a subsequent pregnancy6 . Over the last 10 years there
have been multiple attempts to study the prevalence and
clinical significance of apparently ‘defective’ CS scars
visualized using ultrasonography7 – 14 . In these studies
the prevalence of visible CS scars ranged from 19% to
88% and, despite similar imaging protocols, there was
no agreement regarding the definition of apparent scar
‘defects’.
Despite this interest in CS scar morphology, there have
been no longitudinal studies to evaluate the potential
relationship between scar appearance on ultrasound
imaging across all three trimesters of pregnancy and
pregnancy outcome. Furthermore, no study has evaluated
whether the rate of change in size of different scar
parameters over the course of pregnancy may predict
the likelihood of successful VBAC.
In the current study we aimed to use single measurements of CS scar dimensions in the second trimester of
pregnancy, as well as the change in these dimensions from
the first to the second trimester, to predict the likelihood
of success in women selected for a trial of VBAC.
METHODS
This study was conducted at the ultrasound department of
a London university teaching hospital between June 2010
and July 2011. The local Research and Ethics Committee
approved the protocol. Written informed consent was
obtained from all participating women. In total, 320
women with a singleton pregnancy were recruited; we
used this cohort in our previous publication to develop
reference data relating to the changes in CS scar size seen
over the course of pregnancy15 . All women in this cohort
who had a history of one previous CS were considered
suitable for a trial of VBAC and were referred to a
specialized VBAC clinic, where appropriate counseling
was offered. Women with two or more previous CS were
booked for an elective CS, as per hospital policy.
Women were scanned transvaginally by one of two

operators (O.N. and A.Sm.), using a Voluson E8
Expert (GE Healthcare Ultrasound, Milwaukee, WI,
USA) ultrasound system equipped with a 5–9-MHz
transvaginal probe, at 11–13, 19–21 and 32–34 weeks’
gestation. This coincided with their routine scans
for measurement of fetal nuchal translucency thickness, anomaly screening and, when indicated, growth
assessment. At the first scan, demographic data were
recorded and an obstetric history was obtained. Operators performing the scans were blinded to this clinical
information.
Copyright  2013 ISUOG. Published by John Wiley & Sons Ltd.
Hypoechoic segment
RMT segment
Figure 1 Ultrasound image (sagittal view) of pregnant uterus at
20 weeks, showing both hypoechoic part and residual myometrial
thickness (RMT) of a Cesarean section scar.
A CS scar was defined as being visible when it showed a
hypoechoic indentation representing myometrial discontinuity at the anterior wall of the lower uterine segment,
with or without underlying residual myometrium. The
lower uterine segment was assessed using real-time sonography to identify the area likely to contain the CS scar,
and the delineation and measurement of the scar were carried out according to methods reported previously14 – 16
(Figure 1). The part of the scar contained in any residual myometrium, expressed for measurement purposes
as residual myometrial thickness (RMT) was measured
in the sagittal plane. As a relationship between scar
appearance and function has not been shown, we chose
to avoid using the term ‘defect’ as a morphological
descriptor.
Scar measurements were obtained at each of the three
scan visits, and patients were followed-up until the final
pregnancy outcome was known. Clinicians responsible
for the care of these women in labor were blinded to
the ultrasound features of the CS scar. All data were
transmitted weekly to a designated research database and
were checked for missing, out-of-range and inconsistent
values. For the purposes of this study, uterine scar rupture
was defined as a full-thickness separation of the uterine
wall and overlying serosa observed at CS1 .
Statistical analysis
The factors considered for use in the prediction model
were selected on the basis of expert opinion. They
included two demographic variables (patient age and prepregnancy body mass index (BMI)), one variable related
to previous obstetric history (prior successful VBAC) and
two CS scar variables (RMT measured at the second
trimester and change in RMT from first to the second
trimester (!RMT)). A logistic regression model was built.
Because the sample size was small relative to the number
of predictors, penalized maximum likelihood based on the
corrected Akaike’s Information Criterion (AIC) was used
to obtain reliable regression coefficients for prediction17 .
Ultrasound Obstet Gynecol 2013; 41: 672–678.
674
We did not find non-linear predictor effects or outliers, nor
did we detect strong relationships among the predictors.
The area under the receiver–operating characteristics
curve (AUC) and the sensitivity and specificity at a range
of cut-off values for predicted probability were computed
to assess the discriminative performance of the model. A
calibration plot was produced to investigate the agreement
between observed and predicted probabilities.
The clinical utility of the model obtained was assessed
with decision curves that plot the net benefit for a range
of cut-off values for predicted probability18,19 . A cutoff of 0.7 was often used in earlier studies2 . Using this
methodology, the clinical utility of a model at a given
cut-off to classify patients as suitable for vaginal delivery
is quantified as the net benefit, which is calculated as the
difference between the proportion of true positives and
the proportion of false positives weighted by the odds
of the probability cut-off. The odds of the cut-off value
represents the relative misclassification cost, with a value
above 1 indicating that an emergency CS (i.e. a false
positive) is worse than is an unnecessary elective CS (i.e. a
false negative). For example, using a probability cut-off of
0.7 implies a weight of 0.7/0.3. = 2.3, such that avoiding
one emergency CS is equivalent to 2.3 unnecessary elective
CS deliveries.
A model has clinical utility at a given cut-off if
its net benefit is higher than the net benefit of two
default strategies for which no model is needed: in
this case attempting vaginal delivery in all patients or
performing elective CS in all patients. The desired cutoff (of probability of success) in order to attempt vaginal
delivery, and thus the relative importance of false-positives
and false-negatives, is not the same for every clinician
or patient. We therefore investigated the relevant range
between 0.3 and 0.9, and plotted the results in a decision
curve.
To validate the prediction models internally we did not
split the data into training and test sets due to the limited
sample size; instead, the regular bootstrap was used to
estimate performance on new patients20,21 .
Analyses were performed using R (http://www.rproject.org/; R Foundation, Vienna, Austria). The ‘rms’
package was used to fit the logistic regression model,
calculate the penalized coefficients and the AUC and
produce the calibration plot16 .
RESULTS
The CS scar was visible in 284/320 (89%) cases. Of
the 36 women in whom the scar was not visible, 22
had undergone one previous CS and of these 18 (82%)
subsequently had a successful vaginal delivery. As the
‘non-visible scar’ group did not have measurable scar
features, subsequent analysis in this article is focused on
the ‘visible scar’ group (Figure 2).
In the visible scar group, 153 women had a history of
more than one previous CS and were therefore assigned to
have an elective CS at ≥ 39 weeks’ gestation; these women
were excluded from development of the prediction model.
Copyright  2013 ISUOG. Published by John Wiley & Sons Ltd.
Naji et al.
Table 1 Indications for emergency Cesarean section (CS) in 47
women with visible CS scar and one previous CS, who attempted
vaginal birth after Cesarean section
n (%)
Indication
Confirmed scar rupture
Antepartum hemorrhage
Placenta previa
Breech presentation
Fetal distress
Fetal distress secondary to placental abruption
Failure to progress
PET/GDM/PROM
Placental abruption
2 (4.3)
2 (4.3)
2 (4.3)
3 (6.4)
14 (29.8)
1 (2.1)
20 (42.6)
2 (4.3)
1 (2.1)
GDM, gestational diabetes mellitus; PET, pre-eclamptic toxemia;
PROM, premature rupture of membranes.
Women with a history of one previous CS and visible
scar (n = 131) were offered a trial of VBAC. Ten women
were subsequently excluded because they underwent CS
prior to labor due to various causes. Of the remaining
121 cases, successful VBAC was achieved in 74 (61%),
while 47 cases had failed trials of labor that resulted in an
intrapartum emergency CS. The indications for CS in the
failed VBAC group are shown in Table 1. Around 80%
were for fetal distress, failure to progress or possible scar
rupture.
The average birth weight was 3170 g in the cases of
failed VBAC and 3410 g in the successful VBAC group.
The mean gestational age at delivery was 38 + 5 (median,
39) weeks in the failed VBAC group and 40 + 1 (median,
40) weeks in the successful VBAC group. A total of
30/47 (64%) cases in the failed and 57/74 (77%) cases
in the successful VBAC group went into spontaneous
labor, with labor induced in the remaining women. No
differences between these groups were seen in the use
of prostaglandins, although more women with a failed
trial of VBAC received Syntocinon. Descriptive statistics
for other variables according to success or failure of
VBAC are described in Table 2. The median RMT in the
second trimester was higher for patients who subsequently
underwent a successful VBAC than it was for patients with
a failed VBAC (4.2 vs 2.8 mm) and the median difference
in RMT from the first to the second trimester (i.e. !RMT)
was lower for women with a successful VBAC than it was
for women with failed trial of labor (0.8 vs 2.8 mm).
In a logistic regression model including age, BMI,
second-trimester RMT, !RMT and previous VBAC, BMI
had a small, but statistically non-significant, positive
effect on the chance of successful vaginal delivery (odds
ratio (OR) = 1.08, P = 0.23). This effect is contrary to
findings in the literature2 and BMI was excluded from the
prediction model to enhance generalizability.
The fitted logistic regression model suggests that the
chance of a successful VBAC increases with larger secondtrimester RMT (OR = 6.26 per mm, P = 0.0009). In
addition, the greater the decrease in RMT from the first to
the second trimester (!RMT), the lower the probability
of a successful VBAC (OR = 0.25 per mm, P < 0.0001).
Although not statistically significant, the effect of maternal
Ultrasound Obstet Gynecol 2013; 41: 672–678.
675
TVS of Cesarean scar to predict successful vaginal birth
Study population (n = 320)
Visible scars
(n = 284)
≥ 2 previous CS:
Planned CS ≥ 39 weeks
(n = 153)
ERCD
(n = 136)
Non-visible scars
(n = 36)
≥ 2 previous CS:
Planned CS ≥ 39 weeks
(n = 14)
1 previous CS:
Planned VBAC ≥ 38 weeks
(n = 131)
EmRCD < 38 weeks
(n = 10)
EmRCD < 38 weeks
(n = 2)
ERCD
(n = 11)
Failed VBAC
EmRCD > 38 weeks
(n = 47)
Semi-elective CS < 39 weeks
(premature uterine contractions)
(n = 17)
1 previous CS:
Planned VBAC ≥ 38 weeks
(n = 22)
Failed VBAC
EmRCD > 38 weeks
(n = 2)
Semi-elective CS < 39 weeks
(premature uterine contractions)
(n = 3)
Successful VBAC ≥ 38 weeks
(n = 74)
Successful VBAC ≥ 38 weeks
(n = 18)
Figure 2 Flow chart summarizing the study group, showing mode of delivery according to Cesarean section (CS) scar visibility. EmRCD,
emergency repeat Cesarean delivery; ERCD, elective repeat Cesarean delivery; VBAC, vaginal birth after Cesarean section.
Table 2 Descriptive statistics of patient, scar, labor and birth weight characteristics according to mode of delivery in 121 women with visible
Cesarean section (CS) scar and one previous CS
Parameter
Age (years)
Body mass index (kg/m2 )
Previous VBAC
RMT (mm)
First trimester
Second trimester
Third trimester
!RMT: first trimester – second trimester (mm)
GA at delivery (weeks)
Birth weight (g)
Spontaneous onset of labor
Use of Syntocinon
Use of prostaglandin
Failed VBAC (n = 47)
Successful VBAC (n = 74)
33 (20, 28, 35, 43)
25 (19, 23, 29, 40)
3 (6.4)
32 (21.0, 29.0, 34.0, 39.0)
27 (18.0, 24.0, 30.0, 41.0)
25 (33.8)
5.5 (3.0, 4.8, 6.4, 9.1)
2.8 (0.5, 2.6, 3.1, 4.2)
2.5 (0.5, 2.4, 2.6, 3.8)
2.8 (0.1, 2.1, 3.4, 6.2)
39 (31, 38, 40, 42)
3170 (1495, 2926, 3600, 4756)
30 (63.8)
12 (25.5)
9 (19.1)
5.9 (2.8, 3.5, 5.9, 7.6)
4.2 (2.6, 3.0, 4.6, 6.3)
3.6 (3.2, 2.6, 3.9, 5.9)
0.8 (0.1, 0.4, 1.2, 4.7)
40 (38, 40, 41, 42)
3488 (2420, 3125, 3700, 4540)
57 (77.0)
2 (2.7)
8 (10.8)
Data given as median (minimum, 1st quartile, 3rd quartile, maximum) or n (%). GA, gestational age; RMT, residual myometrial thickness;
VBAC, vaginal birth after Cesarean section.
age was negative while the effect of previous VBAC was
positive (OR = 0.70, P = 0.30 and OR = 3.28, P = 0.22,
respectively). The penalized effect estimates are given in
Table 3 alongside the non-penalized estimates; the former
are preferred for making predictions in new patients (an
example is given in the Discussion).
At a cut-off probability of 0.7, the model had values
for validated sensitivity and specificity of 90% (Table 4).
The training set AUC for the model was 0.96 (95% CI,
0.90–0.98) and the internally validated AUC was 0.94
Copyright  2013 ISUOG. Published by John Wiley & Sons Ltd.
(95% CI, 0.89–0.96). In comparison, a reduced model
excluding the scar features had a training set AUC of 0.66
(95% CI, 0.55–0.77) and an internally validated AUC of
0.62 (95% CI, 0.51–0.73).
A calibration plot showed the estimated probabilities
for successful VBAC given by the model to be accurate
(Figure 3). There was some underestimation of the
probability (by 0.1 or less) for predicted probabilities
0.2 and 0.7, but the majority of women have predicted
probabilities outside this range.
Ultrasound Obstet Gynecol 2013; 41: 672–678.
676
Naji et al.
Table 3 Estimated effects of patient and scar characteristics in fitted logistic regression model to predict success of trial of vaginal birth after
Cesarean section
Effect
Intercept
Age (per 5 year increase)
RMT (per mm increase)
!RMT (per mm decrease)
Previous VBAC (yes/no)
βpenalized
βnon-penalized
ORnon-penalized (95% CI)
P
−0.77
−0.29
1.44
−1.22
1.09
−1.34
−0.36
1.83
−1.39
1.19
—
0.70 (0.35–1.37)
6.26 (2.12–18.52)
0.25 (0.13–0.48)
3.28 (0.50–21.47)
0.6256
0.2957
0.0009
< 0.0001
0.2157
OR, odds ratio; RMT, second trimester residual myometrial thickness; !RMT, change in RMT from first to the second trimester; VBAC,
vaginal birth after Cesarean section.
Table 4 Sensitivity and specificity for success of trial of vaginal
birth after Cesarean section using the prediction model at specified
cut-off values for predicted probability
0.3
0.5
0.7
0.9
Sensitivity
Specificity
97 (97)
95 (94)
90 (90)
59 (59)
74 (74)
83 (81)
91 (90)
98 (96)
Sensitivity and specificity are given as percentages. Values validated
by a bootstrap procedure are shown in parentheses.
0.8
Net benefit
Cut-off for probability
1.0
0.6
0.4
0.2
1.0
0.0
Actual probability
0.8
0.3
0.4
0.5
0.6
0.7
Threshold probability
0.8
0.9
0.6
Figure 4 Decision curves for three strategies in the management of
delivery in women with visible Cesarean section (CS) scar and one
previous CS: using prediction model for success of vaginal birth
), giving all women a trial of
after Cesarean section (VBAC) (
) and giving no woman a trial of VBAC (
).
VBAC (
Validated net benefit, based on probability cut-offs of 0.3, 0.5, 0.7
and 0.9, is also shown ( ).
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
Figure 3 Calibration plot for model to predict success of vaginal
birth after Cesarean section (VBAC), based on patient age, previous
history of VBAC, residual myometrial thickness (RMT) and change
in RMT from first to second trimester, in 121 women with visible
),
Cesarean section (CS) scar and one previous CS: apparent (
) and ideal (
) calibration. Vertical lines at top
validated (
represent individual data points.
successful VBACs per 100 patients at the same rate of
emergency CS. The model also has clinical utility compared to the default strategy of giving all women a trial of
VBAC. This was even the case at a cut-off of 0.3, at which
the validated net benefit of the model was 0.11 higher.
This is equivalent to 26 (= 0.11 × 100/(0.30/0.70)) fewer
emergency CS deliveries per 100 patients with no increase
in the number of unnecessary elective CS deliveries.
Clinical usefulness
DISCUSSION
The decision curve for our prediction model is shown
in Figure 4. The validated net benefit, based only on
cut-offs of 0.3, 0.5, 0.7 and 0.9, are also shown. The
model clearly shows greater net benefit across the range
of probability thresholds considered, compared with the
default strategies. At the cut-off probability of 0.7, the
validated net benefit of the model was 0.45 higher than
was that of giving none of the women a trial of VBAC.
The interpretation of this value is that the benefit of using
the model with scar characteristics in addition to patient
characteristics is equivalent to identifying 45 additional
We have shown that when sonographic measurements of
CS scars during pregnancy, namely RMT in the second
trimester and change in RMT from the first to the second
trimester, are incorporated into prediction models, the
success of a trial of labor can be predicted with a high level
of accuracy. Previously published models have focused on
single RMT measurements made in the late third trimester.
In contrast, our study suggests that if an assessment of the
scar is made relatively early in pregnancy then decisions
regarding the mode of delivery can be made well in
advance of any potential delivery date, hence our decision
Predicted probability
Copyright  2013 ISUOG. Published by John Wiley & Sons Ltd.
Ultrasound Obstet Gynecol 2013; 41: 672–678.
TVS of Cesarean scar to predict successful vaginal birth
to examine changes in RMT from the first to the second
trimester rather than from the second to the third.
We found that women who underwent successful trials
of labor had significantly greater RMT measurements in
the second trimester in comparison with women who had
a failed VBAC. Furthermore, women with a successful
VBAC showed minimal changes in RMT from the first
to the second trimester. It is interesting that it is the
RMT that seems predictive of outcome rather than the
size of the hypoechoic part of the scar (the so-called
niche or ‘defect’). We have previously suggested that,
even in the presence of a large hypoechoic segment, the
scar is not necessarily defective as long as the RMT is
sufficiently thick and remains so during the pregnancy15 .
Moreover, the outcome when the scar could not be
visualized perhaps supports this, as when a trial of VBAC
was attempted in these women, 82% of cases resulted in
a successful vaginal delivery. We think that an inability
to visualize a scar (providing the assumed location of the
CS scar can be located anatomically) probably reflects
good scar healing, with no hypoechoic segment on
ultrasound. This issue is important, as anecdotally we
have noticed a trend for clinicians to request assessments
of CS scars using ultrasound to determine whether they
are ‘deficient’. Our findings suggest that the observation of
a large hypoechoic part of the scar may not indicate that
scar integrity is compromised, providing that the RMT
is adequate.
There have been several reports describing prediction
models to help clinicians counsel women regarding
the mode of delivery after previous CS. However, to
our knowledge, the use of a combination of CS scar
measurements with clinical variables to predict the
success or failure of a trial of VBAC has not been
described previously. Studies published to date have
focused on either single RMT measurements at 36 weeks’
gestation22 – 24 or solely on demographic and clinical
factors2 . Eden et al.2 carried out a systematic review
to evaluate VBAC prediction models and concluded that
none consistently identified women at risk of a failed
trial of labor. We have introduced the change in RMT
over time as a predictive variable. As the uterus changes
in size so rapidly during the course of a pregnancy, we
felt it likely that the speed of change in scar dimensions
may reflect scar integrity. Our data suggest that this may
indeed be the case.
The studies included in the review by Eden et al.2
were mainly retrospective cohort and case–control
studies. Whilst previously reported models contain both
antepartum and intrapartum variables, none has included
information relating to the appearance of the CS scar.
Including scar measurements in our model enabled us
to predict success of VBAC based on a small number
of relatively easily obtained variables. For example, at a
0.7 cut-off, the sensitivity of the reduced model (patient
characteristics only) was 33%, but this improved to 90%
when CS scar features were added (full model); this could
limit the need to include an exhaustive list of clinical
variables.
Copyright  2013 ISUOG. Published by John Wiley & Sons Ltd.
677
A strength of our model is that it provides both
patient and clinician with a calibrated probability for
the likely success of a trial of VBAC. We can illustrate
this by giving two relatively extreme examples. The
formula for the probability of success (based on the
coefficients provided in Table 2) is P = 1/(1 + exp(−z)),
where ‘exp’ refers to the exponential function and
z = −0.77− 0.29 × age/5 + 1.44 × RMT−1.22 × !RMT
+ 1.09 × (1 if previous VBAC). If we consider a woman
aged 32, without a previous VBAC, with an RMT on her
second-trimester scan of 2.7 mm and a decrease in RMT
of 1.5 mm from the first to the second trimester, then,
by entering the specified values, the estimated chance of
successful vaginal delivery for this patient is 36%. In
contrast, if the same woman had an RMT of 4.1 mm at
the second-trimester ultrasound scan and a decrease in
RMT of 0.9 mm from the first to the second trimester, she
would have an estimated 90% chance of a successful
vaginal delivery. There were two cases of confirmed
uterine rupture in the group selected for a trial of VBAC
in our study population. If we apply the model in these
cases, the estimated chance of their achieving a successful
VBAC would be 0.00005% and 0.00009%, respectively.
Had the model been applied prospectively, both women
would have been strongly advised against a trial of VBAC
and so would have avoided the morbidity associated with
uterine rupture.
The AUC for our model is perhaps surprisingly high.
This may reflect the relatively small number of women
in the study. However, even if interpreted with caution,
the findings suggest that the scar is a major factor in the
function of the uterus during a trial of labor. The main
indications for repeat CS in the failed VBAC group were
fetal distress and failure to progress. It is entirely plausible
that both these factors could be exacerbated by poor scar
healing. Furthermore, a poorly healed scar may impact
on decidualization and subsequent placentation and be
associated with a greater likelihood of fetal distress in
labor.
A key issue for any test is reproducibility. We have
shown previously that the interobserver variability for
measurement of CS scars is low in the first trimester
and moderate in the second and third trimesters14 . It is
therefore reasonable to suggest that these measurements
will be of value if used more widely. It is important
to highlight that whilst this model and approach show
promise, external validation in different units and
populations is needed before introduction into clinical
practice.
CS rates have increased globally over recent years and,
as a result, the number of women approaching their
second pregnancy with a scar on the uterus is also
increasing. Consequently, establishing reliable ways in
which to advise this group of women on the likelihood
of their succeeding in a future trial of labor is of
increasing importance. We have shown that the inclusion
of sonographic RMT measurements in a prediction model
can give an accurate quantitative assessment of the
likelihood of successful vaginal delivery after a previous
Ultrasound Obstet Gynecol 2013; 41: 672–678.
678
CS. If the observed test performance is retained on external
validation, this ‘scar-based’ approach to advising women
contemplating a trial of VBAC could allow them to
make well-informed decisions regarding their options for
delivery relatively early in the course of pregnancy.
ACKNOWLEDGMENTS
T. Bourne and S. Ghaem-Maghami are supported by
the National Institute for Health Research (NIHR),
Biomedical Research Centre based at Imperial College
Healthcare NHS Trust and Imperial College London.
The views expressed are those of the author(s) and not
necessarily those of the NHS, the NIHR or the Department
of Health.
L. Wynants is supported by a PhD fellowship from
the Flanders’ Agency for Innovation by Science and
Technology (IWT-Vlaanderen), by the Research Council
KU Leuven (GOA MaNet), by the Flemish Government
(iMinds) and by the Belgian Federal Science Policy Office
(IUAP P7/DYSCO). B. Van Calster is supported by a
postdoctoral fellowship from the Research Foundation
Flanders (FWO).
REFERENCES
1. Naji O, Abdallah Y, Paterson-Brown S. Cesarean birth: surgical techniques. Glob Libr Women’s Med 2010; DOI: 10.3843/
GLOWM.10133. http://www.glowm.com/section_view/item/
133.
2. Eden KB, McDonagh M, Denman MA, Marshall N, Emeis C,
Fu R, Janik R, Walker M, Guise JM. New insights on vaginal
birth after cesarean: can it be predicted? Obstet Gynecol 2010;
116: 967–981.
3. Flamm BL, Geiger AM. Vaginal birth after cesarean delivery: an
admission scoring system. Obstet Gynecol 1997; 90: 907–910.
4. Dinsmoor MJ, Brock EL. Predicting failed trial of labor
after primary cesarean delivery. Obstet Gynecol 2004; 103:
282–286.
5. McMahon MJ, Luther ER, Bowes WA, Jr, Olshan AF.
Comparison of a trial of labor with an elective second cesarean
section. N Engl J Med 1996; 335: 689–695.
6. Ozdemir I, Yucel N, Yucel O. Rupture of the pregnant uterus:
a 9-year review. Arch Gynecol Obstet 2005; 272: 229–231.
7. Ofili-Yebovi D, Ben-Nagi J, Sawyer E, Yazbek J, Lee C,
Gonzalez J, Jurkovic D. Deficient lower-segment Cesarean
section scars: prevalence and risk factors. Ultrasound Obstet
Gynecol 2008; 31: 72–77.
8. Bij de Vaate AJ, Brolmann HA, van der Voet LF, van der
Slikke JW, Veersema S, Huirne JA. Ultrasound evaluation of
the Cesarean scar: relation between a niche and postmenstrual
spotting. Ultrasound Obstet Gynecol 2011; 37: 93–99.
9. Osser OV, Jokubkiene L, Valentin L. High prevalence of
defects in Cesarean section scars at transvaginal ultrasound
examination. Ultrasound Obstet Gynecol 2009; 34: 90–97.
Copyright  2013 ISUOG. Published by John Wiley & Sons Ltd.
Naji et al.
10. Wang CB, Chiu WW, Lee CY, Sun YL, Lin YH, Tseng CJ.
Cesarean scar defect: correlation between Cesarean section
number, defect size, clinical symptoms and uterine position.
Ultrasound Obstet Gynecol 2009; 34: 85–89.
11. Menada Valenzano M, Lijoi D, Mistrangelo E, Costantini S,
Ragni N. Vaginal ultrasonographic and hysterosonographic
evaluation of the low transverse incision after caesarean section:
correlation with gynaecological symptoms. Gynecol Obstet
Invest 2006; 61: 216–222.
12. Regnard C, Nosbusch M, Fellemans C, Benali N, van
Rysselberghe M, Barlow P, Rozenberg S. Cesarean section scar
evaluation by saline contrast sonohysterography. Ultrasound
Obstet Gynecol 2004; 23: 289–292.
13. Armstrong V, Hansen WF, Van Voorhis BJ, Syrop CH.
Detection of cesarean scars by transvaginal ultrasound. Obstet
Gynecol 2003; 101: 61–65.
14. Naji O, Daemen A, Smith A, Abdallah Y, Saso S, Stalder
C, Sayasneh A, McIndoe A, Ghaem-Maghami S, Timmerman
D, Bourne T. Visibility and measurement of cesarean section
scars in pregnancy: a reproducibility study. Ultrasound Obstet
Gynecol 2012; 40: 549–556.
15. Naji O, Daemen A, Smith A, Abdallah Y, Saso S, Stalder C,
Sayasneh A, McIndoe A, Ghaem-Maghami S, Timmerman D,
Bourne T. Changes in Cesarean section scar dimensions during
pregnancy: a prospective longitudinal study. Ultrasound Obstet
Gynecol 2013; 41: 556–562.
16. Naji O, Abdallah Y, Bij De Vaate AJ, Smith A, Pexsters A,
Stalder C, McIndoe A, Ghaem-Maghami S, Lees C, Brolmann
HA, Huirne JA, Timmerman D, Bourne T. Standardized
approach for imaging and measuring Cesarean section scars
using ultrasonography. Ultrasound Obstet Gynecol 2012; 39:
252–259.
17. Harrell FE, Jr. Regression Modeling Strategies With Applications to Linear Models, Logistic Regression, and Survival
Analysis. Springer: New York, 2001.
18. Vickers AJ. Prediction models in cancer care. CA Cancer J Clin
2011 Jun 23. DOI: 10.3322/caac.20118.
19. Vickers AJ, Elkin EB. Decision curve analysis: a novel method
for evaluating prediction models. Med Decis Making 2006; 26:
565–574.
20. Steyerberg EW. Clinical Prediction Models. A practical
Approach to Development, Validation, and Updating. Springer:
New York, 2010.
21. Steyerberg EW, Harrell FE, Jr, Borsboom GJ, Eijkemans MJ,
Vergouwe Y, Habbema JD. Internal validation of prediction
models: efficiency of some procedures for logistic regression
analysis. J Clin Epidemiol 2001; 54: 774–781.
22. Martins WP, Barra DA, Gallarreta FM, Nastri CO, Filho
FM. Lower uterine segment thickness measurement in
pregnant women with previous Cesarean section: reliability
analysis using two- and three-dimensional transabdominal and
transvaginal ultrasound. Ultrasound Obstet Gynecol 2009; 33:
301–306.
23. Jastrow N, Chaillet N, Roberge S, Morency AM, Lacasse Y,
Bujold E. Sonographic lower uterine segment thickness and risk
of uterine scar defect: a systematic review. J Obstet Gynaecol
Can 2010; 32: 321–327.
24. Jastrow N, Antonelli E, Robyr R, Irion O, Boulvain M. Interand intraobserver variability in sonographic measurement of
the lower uterine segment after a previous Cesarean section.
Ultrasound Obstet Gynecol 2006; 27: 420–424.
Ultrasound Obstet Gynecol 2013; 41: 672–678.